Background: With the introduction of new anti-tuberculosis drugs, all-oral regimens with shorter treatment durations for multidrug-resistant tuberculosis have been anticipated. We aimed to investigate whether a new all-oral regimen was non-inferior to the conventional regimen including second-line anti-tuberculosis drugs for 20–24 months in the treatment of fluoroquinolone-sensitive multidrug-resistant tuberculosis. Methods: In this multicentre, randomised, open-label phase 2/3 non-inferiority trial, we enrolled men and women aged 19–85 years with multidrug-resistant tuberculosis confirmed by phenotypic or genotypic drug susceptibility tests or rifampicin-resistant tuberculosis by genotypic tests at 12 participating hospitals throughout South Korea. Participants with fluoroquinolone-resistant multidrug-resistant tuberculosis were excluded. Participants were randomly assigned (1:1) to two groups using a block randomisation, stratified by the presence of diabetes and cavitation on baseline chest radiographs. The investigational group received delamanid, linezolid, levofloxacin, and pyrazinamide for 9 months, and the control group received a conventional 20–24-month regimen, according to the 2014 WHO guidelines. The primary outcome was the treatment success rate at 24 months after treatment initiation in the modified intention-to-treat population and the per-protocol population. Participants who were “cured” and “treatment completed” were defined as treatment success following the 2014 WHO guidelines. Non-inferiority was confirmed if the lower limit of a 97·5% one-sided CI of the difference between the groups was greater than −10%. Safety data were collected for 24 months in participants who received a predefined regimen at least once. This study is registered with ClinicalTrials.gov, NCT02619994. Findings: Between March 4, 2016, and Sept 14, 2019, 214 participants were enrolled, 168 (78·5%) of whom were included in the modified intention-to-treat population. At 24 months after treatment initiation, 60 (70·6%) of 85 participants in the control group had treatment success, as did 54 (75·0%) of 72 participants in the shorter-regimen group (between-group difference 4·4% [97·5% one-sided CI –9·5% to ∞]), satisfying the predefined non-inferiority margin. No difference in safety outcomes was identified between the control group and the shorter-regimen group. Interpretation: 9-month treatment with oral delamanid, linezolid, levofloxacin, and pyrazinamide could represent a new treatment option for participants with fluoroquinolone-sensitive multidrug-resistant tuberculosis.